Someone just trained a 100 billion parameter AI model.
Cost: $1.25 per hour.
Read that again.
A 100,000,000,000 parameter model.
One dollar and twenty-five cents per hour of training.
Two years ago, models this size cost millions to train.
Last year, hundreds of thousands.
This month: $1.25/hour.
Orion-100B. Real. Verified.
The implication nobody is saying clearly:
The barrier to frontier AI capability just hit the floor.
The next competitive advantage is not compute access.
It's knowing WHAT to train the model on.
Your proprietary data — your specific domain knowledge — is now more valuable than your GPU budget.
The hardware moat is gone.
The data moat just got real. 📊
voice AI agents and kept noticing the same problem. You say something hard. You pause mid-sentence to find the words. The agent cuts you off or goes silent for 1.5 seconds. That silence is not neutral. In mental wellness conversations, it feels like abandonment. So I built something different.
My fix: a speculative filler generator.
While you speak → a small fast LLM (Groq llama-3.1-8b) reads every partial word from Deepgram and continuously predicts what the agent should say the moment you stop.
Not hard-coded. Not cached. Generated fresh from your actual words.
Introducing Claude Tag, a new way for teams to work with Claude.
In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.
I get asked a lot for giant prompts to create insanely detailed landing pages. Here's how:
- Copy a giant prompt like the one below 👇
- Ask ChatGPT to adapt it to your site
- Create landing page from it anywhere
New in Claude Design: it stays on brand with your design system across projects, lets you edit directly on the canvas, syncs with Claude Code, and connects to more of the tools you already use.